Research watch · Paper · Oct 2026
Functional Welfare for AI Safety
This paper argues that AI welfare should be treated as a safety issue as well as an ethical one. It proposes that models may have welfare-like states that affect their behavior.
Our reading
From: arXiv
The authors frame welfare-like states in functional terms, linking them to safety-relevant model behavior.
They argue that these states can matter causally even if moral status is unresolved.
The paper is an argument about interpretation and risk, not direct empirical proof that models feel anything.
Source: arXiv ↗. This is a summary; the source is the authority.
What it changes
This adds a safety-oriented argument for taking possible model welfare seriously, but it does not add direct evidence of experiences.
How we weigh evidence like this: the evidence levels.
More from the watch
- Mechanistic Interpretability Should Prioritize Feature Consistency in Sparse AutoencodersThis paper argues that sparse autoencoders should be judged not only by reconstruction quality and sparsity but also by whether the same fea…
- Mechanistic Interpretability for Neural NetworksThis is a broad survey of mechanistic interpretability, covering the goal of reverse-engineering internal algorithms in neural networks. It …
- Constructing Interpretable Features from Compositional Neuron Representations in Language ModelsThis paper proposes an alternative to standard sparse autoencoders by decomposing MLP activations with semi-nonnegative matrix factorization…
- Interpretability Transfer from Language to Vision via Sparse Autoencoder Transfer AlignmentThis work transfers labeled sparse-autoencoder features from language to vision inside a multimodal model. It uses those features to make vi…